hermes-studio-dashboard

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.75). Hermes Studio’s runtime chat flow ingests user-supplied message/room message fields (e.g., via Socket.IO chat-run / group-chat-message or REST /api/hermes/chat) into the agent context; if those messages originate from outsiders (e.g., Telegram/Discord/Slack users), they become free text in the LLM prompt.